added use bbox switcher

src/main.py

@@ -20,6 +20,7 @@
 from supervisely._utils import rand_str, is_debug_with_sly_net
 from supervisely.app.content import get_data_dir
 import supervisely.app.development as sly_app_development
+from supervisely.app.widgets import Switch, Field
 from sam2.build_sam import build_sam2, build_sam2_video_predictor
 from sam2.sam2_image_predictor import SAM2ImagePredictor
 from sam2.automatic_mask_generator import SAM2AutomaticMaskGenerator
@@ -40,6 +41,20 @@
 
 class SegmentAnything2(sly.nn.inference.PromptableSegmentation):
 
+    def add_content_to_pretrained_tab(self, gui):
+        self.use_bbox = Switch(switched=True)
+        use_bbox_field = Field(
+            content=self.use_bbox,
+            title="Use bounding box prompt",
+            description=(
+                "Define whether to use bounding box prompt when labeling images and videos or not. "
+                "If turned off, then only point prompts (positive and negative clicks) will be used. "
+                "Adding bounding box prompt can be useful when labeling entire objects, while using "
+                "only point prompts can be better when segmenting specific parts of objects."
+            ),
+        )
+        return use_bbox_field
+
     def support_custom_models(self):
         return False
 
@@ -364,20 +379,21 @@ def generate_artificial_prompt(self, bitmap):
             col_index += origin_col
             point_coordinates.append([col_index, row_index])
             point_labels.append(1)
-        # generate box prompt
-        rectangle = bitmap.to_bbox()
-        padding = 0.03
-        bbox = [
-            rectangle.left * (1 - padding),
-            rectangle.top * (1 - padding),
-            rectangle.right * (1 + padding),
-            rectangle.bottom * (1 + padding),
-        ]
         prompt = {
             "point_coordinates": point_coordinates,
             "point_labels": point_labels,
-            "bbox": bbox,
         }
+        if self.use_bbox.is_switched():
+            # generate box prompt
+            rectangle = bitmap.to_bbox()
+            padding = 0.03
+            bbox = [
+                rectangle.left * (1 - padding),
+                rectangle.top * (1 - padding),
+                rectangle.right * (1 + padding),
+                rectangle.bottom * (1 + padding),
+            ]
+            prompt["bbox"] = bbox
         return prompt
 
     def serve(self):
@@ -492,7 +508,10 @@ def smart_segmentation(response: Response, request: Request):
             try:
                 # predict
                 logger.debug("Preparing settings for inference request...")
-                settings["mode"] = "combined"
+                if self.use_bbox.is_switched():
+                    settings["mode"] = "combined"
+                else:
+                    settings["mode"] = "points"
                 if "image_id" in smtool_state:
                     settings["input_image_id"] = smtool_state["image_id"]
                 elif "video" in smtool_state:
@@ -736,15 +755,16 @@ def track(request: Request):
                     try:
                         bbox_str = bitmap_frame_data[bitmap_center_str]
                         figure_prompt = frame_prompts[bbox_str]
-                        top, left, bottom, right = bbox_str.split("-")
-                        padding = 0.03
-                        bbox = [
-                            int(left) * (1 - padding),
-                            int(top) * (1 - padding),
-                            int(right) * (1 + padding),
-                            int(bottom) * (1 + padding),
-                        ]
-                        figure_prompt["bbox"] = bbox
+                        if self.use_bbox.is_switched():
+                            top, left, bottom, right = bbox_str.split("-")
+                            padding = 0.03
+                            bbox = [
+                                int(left) * (1 - padding),
+                                int(top) * (1 - padding),
+                                int(right) * (1 + padding),
+                                int(bottom) * (1 + padding),
+                            ]
+                            figure_prompt["bbox"] = bbox
                     except Exception:
                         mode = "artificial clicks"
                 if mode == "artificial clicks":
@@ -764,7 +784,7 @@ def track(request: Request):
                 fig_prompt = figure_data["figure_prompt"]
                 point_coordinates = fig_prompt["point_coordinates"]
                 point_labels = fig_prompt["point_labels"]
-                bbox = fig_prompt["bbox"]
+                bbox = fig_prompt.get("bbox")
                 _, out_obj_ids, out_mask_logits = video_predictor.add_new_points_or_box(
                     inference_state=inference_state,
                     frame_idx=0,